为什么值得关注

它通过新颖的 SSR+ 训练范式，将之前的空间理解模型扩展为完整的感知-规划-行动-评估循环，该范式结合了任务向量合并与定向微调，在不产生跨任务干扰的情况下统一了多种机器人能力。

适合谁使用

具身 AI 研究人员
机器人工程师
自主系统开发人员
AI 安全与监控研究人员

典型使用场景

动态环境中的长程任务规划
机器人导航与操作动作生成
执行进度估计以进行错误恢复
通过回滚反馈实现超越静态模仿的自我改进

项目优势

单个 80 亿参数骨干模型涵盖五种认知功能
在保持广泛空间推理能力的同时扩展到规划与动作生成
SSR+ 训练范式减少了接地、导航、操作和进度估计之间的干扰
支持可执行的导航决策和连续操作控制

使用前须知

为统一覆盖而牺牲了狭窄任务的专门化，可能在个别基准上表现不足
80 亿参数规模对于边缘部署可能计算负担较重
目前仅发布了一个检查点和技术报告，社区采用尚处于初期阶段

README 快速开始

ACE-Brain-0.5: A Unified Embodied Foundational Model for Physical Agentic AI

📑 Contents

🚀 News

2026/07/06: 🔥 We release our technical report and ckpt.

🧠 Introduction

ACE-Brain-0.5 is a unified embodied foundation model for Physical Agentic AI. It extends ACE-Brain-0 from an understanding-centric spatial model into a closed-loop embodied model that can perceive the physical world, plan under goals, act through robot bodies, monitor execution progress, and improve from accumulated experience.

ACE-Brain-0.5 organizes robot intelligence into five tightly coupled cognitive functions: Spatial Perception, Decision Making, Embodied Interaction, Self Monitoring, and Self Improvement. A single 8B backbone instantiates the core perception-planning-action-evaluation loop, supporting object and affordance grounding, 3D and egocentric spatial reasoning, long-horizon task planning, navigation and manipulation action generation, and progress estimation for verification and recovery.

🔥 Key Features

Unified Embodied Foundation Model: Organizes robot intelligence into a single closed-loop model spanning Spatial Perception, Decision Making, Embodied Interaction, Self Monitoring, and Self Improvement.
SSR+ training paradigm: Extends Scaffold-Specialize-Reconcile with a Reactivate stage, combining task-vector merging with targeted fine-tuning to unify spatial reasoning, grounding, navigation, manipulation, and progress estimation without cross-task interference.

🏗️ Method & Architecture

ACE-Brain-0.5 uses a shared embodied backbone to encode heterogeneous inputs and maintain a unified scene-and-task representation, while dedicated interfaces decode this shared state into spatial grounding, executable subgoal planning, navigation and manipulation actions, and progress-estimation signals. Training follows SSR+, which inherits the spatial scaffold from ACE-Brain-0, specializes domain capabilities, reconciles task vectors through model merging, and applies a lig

项目描述

The official repository of ACE-Brain-0.5 unified embodied foundation model.

ACE-Brain-0.5